Policy Optimization via Importance Sampling

Abstract

Policy optimization is an effective reinforcement learning approach to solve continuous control tasks. Recent achievements have shown that alternating online and offline optimization is a successful choice for efficient trajectory reuse. However, deciding when to stop optimizing and collect new trajectories is non-trivial, as it requires to account for the variance of the objective function estimate. In this paper, we propose a novel, model-free, policy search algorithm, POIS, applicable in both action-based and parameter-based settings. We first derive a high-confidence bound for importance sampling estimation; then we define a surrogate objective function, which is optimized offline whenever a new batch of trajectories is collected. Finally, the algorithm is tested on a selection of continuous control tasks, with both linear and deep policies, and compared with state-of-the-art policy optimization methods.

Cited by in corpus (16)

• Off-Policy Policy Gradient with State Distribution Correction
• Sample Efficient Policy Gradient Methods with Recursive Variance Reduction
• POPCORN: Partially Observed Prediction COnstrained ReiNforcement Learning
• Autofocused oracles for model-based design
• An Improved Convergence Analysis of Stochastic Variance-Reduced Policy Gradient
• Black-box Off-policy Estimation for Infinite-Horizon Reinforcement Learning
• How to Learn a Useful Critic? Model-based Action-Gradient-Estimator Policy Optimization
• Confident Off-Policy Evaluation and Selection through Self-Normalized Importance Weighting
• Understanding the Curse of Horizon in Off-Policy Evaluation via Conditional Importance Sampling
• Sparse tree search optimality guarantees in POMDPs with continuous observation spaces
• Efron-Stein PAC-Bayesian Inequalities
• Taylor Expansion Policy Optimization
• Variance-Reduced Off-Policy Memory-Efficient Policy Search
• Parameter-Based Value Functions
• Fault-Tolerant Federated Reinforcement Learning with Theoretical Guarantee
• Conditional Importance Sampling for Off-Policy Learning